The market is shifting. Goldman Sachs just confirmed what many of us in the crypto trenches have been feeling for months: the AI narrative is no longer a single, uniform trade. In July, every AI-linked sector—memory, semiconductors, optical communications, data centers, neocloud—sold off in lockstep, as if they were one giant index fund. But by August, the rebound told a different story. Optical communications bounced 32%, neocloud 20%, AI data centers 17%, while memory limped to 12% and AI power barely moved 6%. The divergence is not noise. It’s a signal. The market is beginning to differentiate between profit cycles, valuations, and fundamentals. And this signal is not just for Wall Street. It’s a mirror for crypto’s own AI obsession.
I’ve been auditing crypto projects since the ICO bubble of 2017. Back then, every whitepaper promised a “decentralized AI” that would revolutionize supply chains or predict election outcomes. Most were vaporware. But the market didn’t care—it bid up tokens based on the label alone. sound familiar? Today, we have dozens of “AI + blockchain” tokens. Some are building inference markets, others are training models on-chain, and a few are just repackaging ChatGPT wrappers with a token. The problem is that the market has treated them all as a single basket. Just like Goldman Sachs observed in traditional AI, the crypto AI trade is now diverging. The question is: which projects have real economic value, and which are just riding the narrative?
Goldman Sachs analysts pointed out that the AI trade phase is not over, but the era of achieving a unified valuation premium solely based on the AI label is ending. This is exactly what we need to internalize in crypto. The “AI” label has been a liquidity magnet for the past year. Projects launching with “AI” in their name raised millions without a working product. But the market is waking up. The same divergence we saw in August—optical communications beating memory—is playing out in crypto. The inference economy, where AI models are run and queried, is emerging as a new mainline. Software and services that enable low-cost inference, like decentralized compute networks, are gaining traction. Meanwhile, memory tokens that rely on the narrative of “AI needs more storage” are stagnating because the market is now asking: “Is the storage actually being used? Or is it just a story?”
Verify the code, trust the community. That’s my signature for a reason. In crypto, the first principle is that code is the ultimate arbiter of value. But when it comes to AI, the code is often opaque. Most AI tokens are not actually on-chain—they are off-chain models with a token on top. The real test is whether the protocol has a defensible moat in the AI value chain. Is it providing compute? Is it hosting inference? Is it verifying model outputs? Or is it just minting tokens to pay for cloud APIs? The market is starting to penalize the latter.
Let me give you a concrete example from my own experience. In 2020, during DeFi Summer, I left a blockchain analytics firm because I saw protocols exploiting users through opaque incentive structures. The same pattern is repeating in AI. Projects are issuing tokens to fund development, but the tokens themselves have no utility beyond speculation. The “AI” label is the new “yield farming.” It’s a narrative that attracts capital, but it doesn’t create sustainable value. The market is now differentiating between projects that have genuine revenue—like those renting out GPU compute to AI researchers—and those that are still burning cash on marketing.
Bulls react. Bears reflect. We build. The current bear market in crypto (and the correction in AI stocks) is a time for reflection. The projects that will survive are those that have a clear value proposition beyond the hype. In the inference economy, the winners will be those that can offer cheap, reliable, and verifiable compute. This is where blockchain can truly shine—by creating decentralized markets for compute that are trustless and transparent. But the current crop of “AI blockchains” often fail because they are centralized under the hood. The multi-sig admin of a smart contract can upgrade the code at any time, which defeats the purpose of trustlessness. This is the same issue I’ve seen in DAO governance: “code is law” is a myth when a few people hold the keys.
Tech changes. Values remain. The core value of blockchain is sovereignty—the ability for individuals to control their own assets and data. AI should be the same. We need AI that is not controlled by a few corporations, but that is open, verifiable, and owned by the community. The tokens that will survive are those that embody this value. Not those that simply say “AI” in their whitepaper.
Now, let’s look at the data. Over the past 30 days, the top 10 AI tokens by market cap have seen a median drop of 15% in trading volume. Meanwhile, decentralized compute networks like Akash and Render have seen their utilization rates increase by 20%. This is the divergence. The market is punishing tokens that are pure speculation and rewarding those that have actual usage. The same pattern that Goldman Sachs observed in optical communications vs. memory is playing out: compute (the “inference” layer) is outperforming storage (the “memory” layer). Because the market is realizing that AI inference is the bottleneck, not storage.
Contrarian angle: Many people think that the AI token bubble will burst completely. I disagree. The bubble is deflating, but the underlying technology is real. The contrarian view is that the best time to invest in crypto AI is now, during the divergence, when prices are more rational. But you have to be selective. The era of buying any token with “AI” in the name is over. Now, you need to analyze the protocol’s economic model. Is the token used to pay for compute? Is there a burn mechanism? Is the team transparent about their node operators? Based on my audit experience, most AI tokens fail the “covenant test”—they have no binding commitment to the community beyond the whitepaper.
Takeaway: The AI trade is not dead. It’s maturing. The market is shifting from a narrative-driven premium to a fundamentals-driven evaluation. For crypto, this means that the projects that will thrive are those that build real infrastructure for the inference economy and that embed the values of decentralization and sovereignty into their code. Don’t just hold. Understand. The era of the blanket AI premium is ending. The era of the covenant is beginning.
Let me walk you through the mechanics. In the inference economy, the value lies in the ability to run AI models at scale. Currently, most inference is done on centralized servers (AWS, Azure, Google Cloud). These are expensive, opaque, and subject to censorship. Decentralized compute networks aim to change that by allowing anyone to rent out their GPU to run models. The token in these networks is used to pay for compute, and it also serves as a stake to ensure honest behavior. This is a classic utility token model, but with a twist: the demand for compute is growing exponentially, so the token price should, in theory, correlate with usage. But in practice, many of these tokens are trading at a premium to their fundamental value because of the AI hype. The divergence we are seeing is the market correcting this premium.
My personal experience: When I launched The Decentralized Mind in 2024, I curated a curriculum that connected zero-knowledge proofs to privacy and AI. I saw firsthand how the educational gap between the technology and the narrative leads to mispricing. Most investors don’t understand the difference between an AI inference protocol and an AI trading bot. They just see “AI” and buy. The same happened in 2017 with ICOs. The market is now learning that the label is not enough.
The soul in the machine. In 2025, I wrote a white paper arguing that without a decentralized ethical framework, AI would consolidate power rather than liberate it. That paper gained traction among EU regulators. Now, I’m seeing the same principle apply to token markets. The market is starting to value projects that have a clear ethical framework—those that are not just extracting value, but building a sustainable ecosystem. The inference economy requires trust. Who verifies the model’s output? Who ensures that the compute provider doesn’t cheat? Blockchain can solve this, but only if the code is designed with covenant in mind.
Final thought: The market is now a testing ground. The projects that can prove their utility, that have a clear revenue model, and that are genuinely decentralized will survive. The rest will fade. This is healthy. It’s the same process that happened in DeFi after the 2020 summer: the yield farms died, but Uniswap, Aave, and Compound thrived. The same will happen in AI. The projects that are building the infrastructure for the inference economy will be the ones that create lasting value. The others will be forgotten.
Verify the code, trust the community. The code of an AI protocol should be transparent, auditable, and immutable. The community should be the ultimate arbiter of the protocol’s direction. If a project has a multi-sig with a few keys, it’s not decentralized. If it claims to be AI but has no on-chain logic, it’s not a blockchain project. The market is waking up to these truths. The test of the bear market will separate the real builders from the narrative riders.
Bulls react. Bears reflect. We build. In this bear market, I’m spending my time auditing the code of AI protocols. I’m looking at their tokenomics, their governance, and their utility. The ones that pass the covenant test will be the ones I recommend to my students. The others will be cautionary tales.
Tech changes. Values remain. The values of sovereignty, transparency, and community are not new. They are the foundation of the crypto movement. The AI narrative is just the latest vehicle for these values. The market divergence is a sign that the vehicle is being inspected. The ones that are built on solid ground will survive. The rest will be scrapped.
In conclusion, the AI trade in crypto is not over. It’s evolving. The era of the blanket premium is ending. The era of the covenant is beginning. And I, for one, am building for the covenant.
